With the current tragic events in Boston coming to a swift close with the 2 perpetrators in custody my thoughts have gone to the storyline of the apparent failure of crowd sourced investigation of both 4Chan and Reddit.  Some people are espousing the notion that crowd sourcing this kind of work should be limited to crowd sourcing data collection - i.e. collecting everyones images and videos.  I think that the 4Chan and Reddit failures are actually not at all definitive on this subject and I think there is plenty of room for crowd sourcing investigation data analysis.

As background I think it is important to look at other tasks that are successfully crowd sourced, while there are many, I would like to key on protein folding.  This effort was keyed by a game that let humans try different protein folding combinations to earn points for successful foldings.  The key to this game being useful is that their was structure given to the participants in the investigation, namely the physics of protein folding was built into the game world.  This key allowed participants who knew nothing about protein folding produce valid folds by simply using the human brain as the first and foremost patten matcher on this planet.  The fact that it was a game may be important, but I am not convinced that gamification is necessary.

The key issue with Reddit and 4Chan was that it lacked what protein folding had: structure.  The investigations had no structure, no constraints, for lack of a better word, no physics to keep participants from going on tangents.  The lack of structure is why I think these types of investigations can still benefit from more crowd sourcing than just data gathering.

Imagine if the FBI had an app that fed users pictures of every frame of video they had and asked the participants to identify unique individuals, and then match known individuals across frames (cross-correlating).  That structure imposed on the problem would turn on and focus millions of people on productive paths and map out individuals movements across the crime scene in a fraction of the time that the limited number of FBI agents could.

Their are a myriad of other analyses that could be accomplished this way, as long as the tasks are structured and the tool (app) being used enables efficient harnessing of non-expert specific analyses (i.e. things all humans are good at, like recognizing faces - Where's Waldo).  Subject matter expert analyses should always be left to experts: i.e. don't ask the public to analyze blood spatter patterns.

The FBI/HSD/Police should invest in the sorts of tools that would enable this kind of analysis, because it would be hugely powerful way of speeding up some important pieces of data analysis.